AI-Assisted Development

AI coding agents can help read documentation, clarify requirements, write examples, and review an implementation. Providing the official pages directly related to the task reduces guesswork and unnecessary context.

Documentation entry points for AI coding agents

This site provides the following plain-text resources:

• llms.txt
• llms-full.txt

Resource	Best use
llms.txt	A documentation index. Use it to find the pages relevant to the current task.
llms-full.txt	The complete documentation combined into one file, primarily for use as an offline reference.
Individual .txt pages	Provide the relevant page directly when the feature, method, or parameter is already known.

Every page has a plain-text version

Every documentation page has a .txt version at the same path. Append .txt to the original URL. For example, the plain-text version of Market Data Introduction is docs/market-data/intro.txt.

For a specific question, use the relevant .txt page. Start with llms.txt when the page location is unknown. For offline work, download llms-full.txt as a complete reference.

Prompt examples

You know what you want to build

I want to implement 〈feature〉 with the Fubon Neo API in 〈programming language〉.

Read 〈documentation host〉/llms.txt and find the directly relevant pages. Then provide the prerequisites, the smallest runnable example, and validation steps. Use only documented API behavior and list anything that remains unclear.

You have an idea but not a plan yet

I have a rough idea for 〈what you want to build〉 with the Fubon Neo API, but I have not chosen an approach.

Use 〈documentation host〉/llms.txt as the basis for our discussion. Outline the feasible options, key technical choices, constraints, and relevant documentation. Confirm the direction with me before proposing the next implementation step.

If the agent cannot read URLs, attach the relevant .txt pages. For offline work, attach llms-full.txt.

Test before live trading

Before live securities trading, use the official test environment to check login, feature integration, and intended operations. It may not cover every production workflow, so confirm what applies before switching to live trading. Download the test environment information package

Let the AI coding agent maintain its skill

Start with the task already in progress. Have the agent complete the work from the official documentation, then ask it to capture the confirmed, reusable lessons in SKILL.md.

Create or update the Neo API skill from this task, the confirmed documentation, and the corrections made:

- Keep only reusable applicability rules, authoritative sources, constraints, and validation steps
- Preserve the original names of API methods, fields, and enum values
- Do not copy large sections of API documentation; keep the source links and conditions for use
- Mark conflicting or unconfirmed information as unresolved
- Move long reference material to references/ and repeatable checks to scripts/

After editing, summarize the rules added, changed, and removed, with their supporting sources.

When the agent chooses the wrong method, misses a constraint, or cannot validate a result, provide the correction and its source, then ask it to update or retire the relevant rule. The skill should retain reusable decisions and validation methods, not accumulate generic reminders.

Keeping information up to date

• Re-read llms.txt and the relevant individual pages at the start of each task.
• Record the download date and documentation or SDK version for an offline llms-full.txt snapshot, and replace it regularly.
• After completing and correcting a task, ask the agent whether the skill needs an evidence-backed update.
• Periodically replay common tasks to confirm that the skill still finds the correct documents, preserves field names, and identifies prerequisites.
• Remove stale, duplicate, or unsupported rules. Current API details remain governed by the latest official documentation.